Re: [PATCH 4/4] arm64: dts: allwinner: h6: Add CPU Operating Performance Points table

From: Maxime Ripard
Date: Fri Feb 15 2019 - 08:53:08 EST


On Thu, Feb 14, 2019 at 10:52:16PM +0800, Frank Lee wrote:
> On Thu, Feb 14, 2019 at 10:38 PM Maxime Ripard
> <maxime.ripard@xxxxxxxxxxx> wrote:
> >
> > Hi,
> >
> > On Thu, Feb 14, 2019 at 08:09:10AM -0500, Yangtao Li wrote:
> > > Add an OPP (Operating Performance Points) table for the CPU cores to
> > > enable DVFS (Dynamic Voltage & Frequency Scaling) on the H6. This
> > > information comes from github.
> > >
> > > Signed-off-by: Yangtao Li <tiny.windzz@xxxxxxxxx>
> > > ---
> > > arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi | 61 ++++++++++++++++++++
> > > 1 file changed, 61 insertions(+)
> > >
> > > diff --git a/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi b/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi
> > > index 57a1390ecdc2..46a4a69eb38f 100644
> > > --- a/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi
> > > +++ b/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi
> > > @@ -28,6 +28,8 @@
> > > enable-method = "psci";
> > > clocks = <&ccu CLK_CPUX>;
> > > clock-latency-ns = <244144>; /* 8 32k periods */
> > > + operating-points-v2 = <&cpu_opp_table>;
> > > + #cooling-cells = <2>;
> > > };
> > >
> > > cpu1: cpu@1 {
> > > @@ -37,6 +39,8 @@
> > > enable-method = "psci";
> > > clocks = <&ccu CLK_CPUX>;
> > > clock-latency-ns = <244144>; /* 8 32k periods */
> > > + operating-points-v2 = <&cpu_opp_table>;
> > > + #cooling-cells = <2>;
> > > };
> > >
> > > cpu2: cpu@2 {
> > > @@ -46,6 +50,8 @@
> > > enable-method = "psci";
> > > clocks = <&ccu CLK_CPUX>;
> > > clock-latency-ns = <244144>; /* 8 32k periods */
> > > + operating-points-v2 = <&cpu_opp_table>;
> > > + #cooling-cells = <2>;
> > > };
> > >
> > > cpu3: cpu@3 {
> > > @@ -55,6 +61,61 @@
> > > enable-method = "psci";
> > > clocks = <&ccu CLK_CPUX>;
> > > clock-latency-ns = <244144>; /* 8 32k periods */
> > > + operating-points-v2 = <&cpu_opp_table>;
> > > + #cooling-cells = <2>;
> > > + };
> > > + };
> > > +
> > > + cpu_opp_table: opp_table {
> > > + compatible = "operating-points-v2";
> > > + opp-shared;
> > > +
> > > + opp@480000000 {
> > > + opp-hz = /bits/ 64 <480000000>;
> > > + opp-microvolt = <800000 800000 880000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> > > + };
> > > +
> > > + opp@720000000 {
> > > + opp-hz = /bits/ 64 <720000000>;
> > > + opp-microvolt = <800000 800000 880000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> > > + };
> > > +
> > > + opp@816000000 {
> > > + opp-hz = /bits/ 64 <816000000>;
> > > + opp-microvolt = <800000 800000 880000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> > > + };
> > > +
> > > + opp@888000000 {
> > > + opp-hz = /bits/ 64 <888000000>;
> > > + opp-microvolt = <800000 800000 940000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> > > + };
> > > +
> > > + opp@1080000000 {
> > > + opp-hz = /bits/ 64 <1080000000>;
> > > + opp-microvolt = <840000 840000 1060000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> > > + };
> > > +
> > > + opp@1320000000 {
> > > + opp-hz = /bits/ 64 <1320000000>;
> > > + opp-microvolt = <900000 900000 1160000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> > > + };
> > > +
> > > + opp@1488000000 {
> > > + opp-hz = /bits/ 64 <1488000000>;
> > > + opp-microvolt = <960000 960000 1160000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> > > + };
> > > +
> > > + opp@1800000000 {
> > > + opp-hz = /bits/ 64 <1800000000>;
> > > + opp-microvolt = <1060000 1060000 1160000>;
> > > + clock-latency-ns = <244144>; /* 8 32k periods */
> >
> > So we definitely want to have that tested, especially since cpufreq
> > can lead to all kind of hard to debug errors (brown-outs, CPU lockups,
> > cache corruption, etc.). I good way to test that would be to use
> > cpufreq-ljt-stress-test here:
> > https://github.com/ssvb/cpuburn-arm/blob/master/cpufreq-ljt-stress-test
> >
> > I'm especially worried about the higher frequencies that will probably
> > make the SoC heat too much
>
> Indeed, in order to avoid this situation, it is best to have cpu cooling
> support(But now it does not support thermal driver? ).
>
> In this case, perhaps we should remove the frequency beyond a certain
> range to avoid the CPU being too hot?

Yeah, that seems like a nice solution until we have the thermal sensor
running.

Maxime

--
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com